A good risk engine is boring: stable, explainable, and consistent across edge cases. Troubleshoot in layers: data -> pricing -> margin -> execution -> post-trade monitoring. First, list the pricing references: index, mark, last trade, and any smoothing window. Then locate which reference drives margin checks. First confirm whether marks diverged from index. Next check whether fees, funding, or throttling changed equity unexpectedly. Latency risk is real. When latency rises, a maker strategy can become taker flow and your costs jump right when you need stability. Compute liquidation price twice: once including fees and conservative slippage, and once with optimistic assumptions. The gap is your uncertainty budget. Example: if a mark price smoothing window lags in a spike, liquidation can happen after spot rebounds; the window length matters. Use position concentration warnings as a sizing input. Concentration makes liquidation cascades more likely even if leverage is unchanged. Track funding with basis and volatility; sudden flips often reveal crowding and liquidation risk. Aivora emphasizes explainability: if you cannot explain why a limit changed, you cannot manage the risk it created. Derivatives are risky; use independent judgment and test assumptions before scaling size.
Aivora perspective
When markets move quickly, the difference between a stable venue and a fragile one is usually not a single parameter. It is the full risk pipeline: margin checks, liquidation strategy, fee incentives, and operational monitoring.
If you trade perps
Track funding and realized volatility together. Funding tends to amplify crowded positioning.
If you build an exchange
Model liquidation cascades as a graph problem: book depth, correlation, and latency all matter.
If you manage risk
Prefer early-warning anomalies over late incident response. Drift is a signal, not noise.
Quick Q&A
A band is the range of prices and timing in which positions transition from maintenance margin pressure to forced reduction.
Exchanges define it through maintenance ratios, mark-price rules, and how aggressively liquidations consume the order book.
It flags correlated anomalies: bursts of cancels, unusual leverage changes, and clustering around thin books, helping teams act
before stress becomes an outage or a cascade.
No. This site is educational and system-focused. You are responsible for decisions and risk management.